Just so people are aware, I am still seeing kswapd runaways (99% cpu usage for a number of hours, causing other processes to go into tight loops as well) with this kernel version (whether or not swap exists) when updatedb runs in the mornings. The machine exports both Samba and NFS shares, and the kernel is configured for high memory (4GB), so the conditions are similar to the previous examples of this behaviour. I remember that someone at some point had an alternative VM implementation that was supposed to correct this problem. Does anyone know who that was and if there are any patches that would apply against 2.4.21-rc1? -- Zed Pobre a.k.a. Zed Pobre PGP key and fingerprint available on finger; encrypted mail welcomed.